基于混合神经网络与有限状态机的区域电网智能告警处理方法研究

Research on Intelligent Alarm Processing Method for Regional Power Grid Based on HNN and FSM

结合混合神经网络(hybrid neural network,HNN)和有限状态机(finite state machine,FSM),开展变电站告警信息处理。首先,为降低HNN模型构建的复杂度,通过纵向上模块化处理、横向上信号归并的信号处理方法简化告警序列。其次,构建HNN权矩阵模型和学习算法,通过对数据样本进行训练和测试,获取线路、母线和变压器三类设备事故和设备异常的逻辑推理及知识表达。然后,开展故障集的关联性分析,构建FSM模型实现信号的排查和告警过程记录,从而形成告警的原因分析和结果处理。最后,通过实际区域电网案例对本算法进行验证,结果证明该方法对于电力系统通用故障告警判断具有快速、容错和学习能力强等特点,对应用于大规模电力系统的在线故障诊断问题的解决具有重要意义。

Combined with hybrid neural network(HNN)and finite state machine(FSM),the substation alarm information is processed.Firstly,in order to reduce the complexity of HNN model construction,the alarm sequence is simplified by the signal processing method of vertical modular processing and horizontal signal merging;secondly,the HNN weight matrix model and learning algorithm are constructed,and the logic reasoning and knowledge expression of accidents and abnormalities of three types of equipment,line,bus and transformer equipment accidents and equipment exceptions,are acquired by training and testing data samples;thirdly,the correlation analysis of fault set is carried out and the FSM model is built to realize signal troubleshooting and alarm process recording,and finally comprehensive analysis results of alarm cause analysis and result processing are formed.The results show that the method has the characteristics of fast,fault tolerance and strong learning ability for the common fault alarm judgement of power system,which is of great significance to solve the online fault diagnosis problem of large-scale power system.